Azure Open AI service returning error : "Public access is disabled. Please configure private endpoint."

Question

Hi Team, I am trying to use azure open ai api service from my databricsk platform and follwoing the quickstart also.
But getting the error like : Public access is disabled. Please configure private endpoint.

But same things is working when i am using chat completion from playground.

What i suppose to do next.

Also do we have to register over limited access form also as we are trying to use only built in deployed model only.
Looking forward to response.

Answer 1

@Chaudhary, Rohit (Contractor) - Thanks for the additional details.

It seems like the Azure Open AI service is not able to access the public network and requires a private endpoint to be configured. To resolve this issue, you need to configure a private endpoint for the Azure Open AI service.

Here are the steps to configure a private endpoint for the Azure Open AI service:

1. Go to the Azure portal and navigate to the resource group where the Azure Open AI service is located.
2. Click on the Azure Open AI service and go to the Networking tab.
3. Click on the Private endpoint connections tab and then click on the + Private endpoint button.
4. In the Add private endpoint blade, select the subscription, virtual network, and subnet where you want to create the private endpoint.
5. Click on the Review + create button and then click on the Create button to create the private endpoint.

Once the private endpoint is created, you can use it to access the Azure Open AI service from your Databricks platform. Regarding your question about registering for limited access form, it depends on the specific Azure Open AI service you are using. If the service requires limited access, then you need to register for it. However, if you are using a built-in deployed model that does not require limited access, then you do not need to register for it. It is also worth noting that the playground chat completion source code may be using a different endpoint or authentication method than the one you are using in your Databricks platform. You may need to update your code to use the private endpoint and authentication method that you have configured for the Azure Open AI service.